Cyborg, Centaur or Chauffeur: How Leaders Should Use AI, with Scott Anthony

Oh, well, that's it. Anyway, let's Believe that wisdom, because you're trying to transfer the wisdom that you learned into students at Tuck Business School right now. What I really want to explore today is, you know, what learning means in today's world, building on what you learned from your experience in business school to really how can we translate that to the workplace? Now, the first question I want to get into is what have you learned about the limits of traditional sage on the stage lecturing? I mean, how useful is the expert recounting the wisdom of theories and models relevant today, and where are there any limits to it?

Well, first, the benefits of it. I think certainly a good model theory framework functions as a set of lenses that allow you to see things that you would otherwise not see and therefore do things that you would otherwise not do. So the underlying ideas that are being transmitted, I think, are equally, if not more important in a world that's hugely volatile and all that. I'd say the sage on the stage model has always had limits. I remember my mentor, Clay Christensen. One of the things he taught me is the half-life of a good educational experience is about a week. So after a week, after hearing something, you've lost half of it. 2 weeks, you've lost another blah, blah, blah, blah, blah. You know the old saying that a lecture is the fastest way to transmit notes from the professor to the students without going through the minds of either. That really rings true. So, you know, of course I think it's really important to communicate ideas, but until people get a chance to actually do something with it, nothing's going to stick. I think that's always been true.

Well, I'm reading a book at the moment by David Hurst on the new ecology of business, and he uses a model which stuck with me. Learning to play golf. You can read a manual which will tell you, keep your head still, keep your left arm straight for as long as you want, but get onto the golf course or to the putting green or to the pitching green, whatever it's called. And that theory is meaningless when you put a club in your hands and you need to try things out. I guess really that's what you're saying about practice being necessary to work.

No. And you know, if you look at the pedagogical approach that a business school or a medical school follows, they're very similar, even though they're working on very different things. What you're trying to do is first you learn about one, then you do one, then you teach one. In medical school, It's working on actual people. In business school, it is the case study methodology where you say, in the course of a 2-year MBA program, you're going to get a chance to simulate 200 to 300 experiences where you can go and act like the protagonist of a case study and work through a problem collectively. It is exactly the equivalent of going from reading about golf to getting on the putting green. You're not playing a tournament. It's not real, but it is a chance to simulate what it's like so you can begin to develop the muscle around it. Yeah. The exact same idea.

Yeah. So, what topics do you find that you're having to develop new approaches to in any way?

Well, I'd say 2 things to that. One is kind of a horizontal, one is a vertical. The horizontal, everybody wants to learn via, through, about AI. So, there is a lot about how do you bring that into a classroom in a way that does not degrade learning, but upgrades it? So this is the things that you are already teaching. How do you use AI as a way to push further, to enable people to experience things differently, and also not lose the basic things they're learning? That's a horizontal. The vertical is also AI. So how do you then get into specific topics around AI? I have a sprint class, as you know, that I just finished teaching called AI and Consultative Decision Making. The basic idea is to simulate what it's like to be a consultant when you're now armed with these amazing tools. You know, it's interesting. When I first taught the class, I thought the class was all about how do we use AI to address some of the known biases in decision-making, groupthink, anchoring effects, and so on. What I've learned from teaching it a few times is, yes, of course you need to do that. And also you need to give students strategies to deal with what are now emerging as some of the limitations of AI. We know, of course, about hallucinations, but There's also idea bubbles and power persuasion and cognitive debt and brain fry. So there's new things emerging that we have to develop techniques to go and counterbalance. You know, AI connects those things, but it's different how you upgrade what you're currently doing and how you do new things as well.

Well, so you've just used a lot of new terms. I heard brain fry, bubbles. Can you, can you just Unpack some of those things, please.

Yeah. So if I, if I go through the list, so idea bubble. So, and a lot of this traces back to research Ethan Mollick, Karim Lakhini, Hilla Lifshitz-Schmitz, and others have done. So idea bubble, the idea is if you're trying to generate ideas using AI, recognize AI systems are all trained on existing knowledge. And because they're all trained on existing knowledge, while you will get really interesting output really quickly, there's a sameness that comes to it. So just like on social media, we can live in filter bubbles where we only see things from people like us. For ideas, if you're using AI to generate new ideas, there's gonna be a bubble of ideas where they all kind of look the same. That's the first one. Power of persuasion. I'm sure, Mark, you've been in a situation like this. You've gotten output from ChatGPT or Claude or Gemini or whatever, and you push back and you say, well, no, that's not quite right. What the systems do then is they double down. And they're really good at pathos, logos, ethos. What Aristotle said way back when are the ways that we make persuasive arguments in convincing you that they're right even when they're wrong. So they kind of metaphorically dig in their heels and try to stick to a current position. Cognitive debt. As we offload more to AI systems, we accumulate a debt in our brain. We don't have the ability to do things that we once had the ability to do. The debt gets too much. We've got real limits. And then finally, brain fry. We are overloaded like never before with information and stimuli. It can be really exciting, but it also can lead to shutdown because it is simply too much. So those are some of the new things that are emerging. There's more than that, of course, but those are the ones that I highlighted in the class.

I'm glad you're doing it. This is— you picked up on some themes that a past podcast guest arrived Erased the idea that obviously generative AI is trained on what models exist. Inductive deductive thinking at best they can come on. David Cropley's published a paper saying that they are 0.25 as good as the very best human idea generators. Simply because of that, they don't have the abductive reason to connect apparently disconnected ideas. Training people up on that is is very helpful. Completely agree with you on that, and certainly. Having the brain fry, it's all too easy to get lazy, which is why I'm inspired from the instruction of, you know, what we said in the last place, the right way is the hard way. You know, it's building muscle, not letting muscle atrophy is the way you do it in the classroom. So have you developed your training programs over time to help that happen?

For sure. And a large piece of it is just grounding students in the understanding, which everybody intuitively gets that the right way is the hard way. And what I do is I borrow a framework again from the work that Malik et al. have done, where based on some of the early research they did, and early means it was done in 2022, '23, so not that long ago, they said you can begin to see different archetypes that emerge as people are using AI systems. You've got one archetype where you're a cyborg, where you fuse things with AI. There's one archetype where you're a centaur, the head of a human, the body of a horse. You task divide. Then there's a third archetype that they call a self-automator. Because I like alliteration, I call it a chauffeur, which is you just get in the back of the car and you let AI drive you wherever you want to go.

Yeah.

So the thing that I tell students is, let's just be conscious of what we're doing. Sometimes you have to get to a destination quickly, and having a chauffeur is great. However, if you're trying to learn new things, a chauffeur is a very dangerous mode to take. You need to learn how to take the wheel and choose whether this is a divide, Mentor or few cyborg tasks. Again, so much of it is just having the consciousness because you will do the easy stuff naturally. As humans, we will default to easy stuff without thinking about it. Then you wake up one day, a professor asks you a follow-up question. You're like, I have no idea. Yeah. Because I've got this beautiful output, but I had nothing to do with creating it. And we've seen a, a number of professors as they try to assess learning are going back to basics. Pen, paper, conversations. That's how you really tell whether people understand. And when you see beautiful output that gets destroyed with a single question, you kind of know what's happened.

Indeed. I was— you're opening up such a Pandora's box of the challenge of bringing this AI into the workplace. And that's just one of the disruptive forces. I mean, obviously the very popular one that we've got at the moment. I was recalling a conversation I had with Ray Dalio. And he was who wrote a great book on AI and the governance of AI. And he recounted a story with a chief executive of Paramount, and he was open on this and said that they'd spent a lot of money bringing Microsoft Copilot into the organization and gave everyone the standard videos. The uptake? Barely moved the video. Passive learning of these challenging concepts Is not great. But then he also recounted a study of a very modest county council. This is a local government organization in the UK, which had invested in customized training for about 100 or so staff. Each had a 4-hour session, and they went from the vast majority of people being negative, fearful, frightened of AI to being engaged, positive, and for it simply by making it the co-creation of design. Any reflections on what that means and what it might mean for training in the workplace?

It certainly resonates with my own quote unquote lived experience. You know, the class that I've been teaching, I ask students at the beginning of class, how facile are they with AI? How much are they using it, et cetera? And at the end of the class, I ask them, have your AI skills improved? And if so, by how much? And all 3 years I've taught it, there has been a significant improvement in skills. More critically, there's been a significant increase of understanding in where the jagged frontier lies. Jagged frontier. Again, credit to Moloch et al. for coming up with the term. The idea is there's not a smooth frontier where we can see exactly what AI can do. It's very jagged because it's incredibly awesome at some things and weirdly bad at others. And the way you figure that out is by dancing at the jagged frontier. So that's what I do in the class. I say, we're not going to read about things. We're not going to watch demo videos about things. I'm gonna create scaffolding. I'm gonna give you constraints. I'm gonna give you a problem to work on, and you're gonna go and do it. And from doing that and pushing them this year, I had a new thing that I've never done before, which is the final output was to create a synthetic consultant. I had no idea what that was. I had no idea how the experiment was gonna go, but by pushing students into areas where no one knew the answer, we all learned collectively and we had greater confidence about where that jagged frontier is. So I had a couple students who said, you know, I always had just used AI to essentially be another search engine or to slightly improve my writing. I've now recognized there's so much more that I can do with it because I have danced near that jagged frontier. I think, as always, learning by doing is the best way to learn with, again, the right guardrails, constraints, scaffolding, support to help maximize the learning that comes.

Loved that phrase, the jagged frontier. It echoes something that I wanted to talk about because one mental model I have in my mind about When, you know, lectures and expertise might be helpful and when experimentation might be helpful is to think about what Cynefin calls complicated versus complex problems. Or think about it as Heifetz and Linsky language, technical challenges where the expert in the room can teach you how to fix the Rolls-Royce engine on a 747 jumbo jet. And if you know that expertise, you can do it. That's where sage on the stage perhaps has a definite role. But there are emerging problems like, where is AI going to be in 12 months' time, 24 months' time? How do we use that in our business? There's no expertise who can really answer that. That's an explore problem. And so you do need a different type of learning and experimentation in the workplace to do it.

For sure. You know, I am a big fan of the Heifetz-Linsky model and, you know, the way I try to simplify So you said technical, the, they would call the other type of challenge an adaptive challenge.

Yeah.

The way that, that I try to simplify what is a pretty complex, not to mix ideas, but complex concept is to say 3 things are generally true of adaptive challenges. Number one, there's no obvious best answer. Number 2, there's a possibility that someone is going to lose. There's gonna be power dynamics shifts, status shifts, and so on. Number 3, there is the certainty of struggle because people have to unlearn before they learn, and unlearning is incredibly uncomfortable. So if we have one of those, then we have to recognize that the process is going to have discomfort in it, and we have to find a way to stay in what Heifetz and Linsky, the zone of productive discomfort for as long as possible. And there's a lot that's then baked into that about how you actually do it. To me, in the classroom, the biggest thing by far is making it a safe environment where people can go and play. And try different things. If people feel safe and they feel like they're having fun, they can tolerate a huge amount of discomfort. If they feel like there's personal risk involved and it feels like work, then things feel really different. It is work, don't get me wrong, but it's fun work. Fun work is just a lot better in my mind for adaptive challenges.

So we've got, I mean, I'd say definitely AI is one of them, but dealing with international trade disruption, I mean, there's so much uncertainty going on in the world. I'm And moving, if I may, from the lessons in the classroom and the business school and what can realistically be done in the workplace, because 2 weeks ago I had a podcast with Professor Joseph Fuller who confirmed with a frightening statistic something that I feared to be true. In the last 25 years, there's been a 50% drop in training budgets in the typical corporation. Most of that remaining money is spent on things like compliance training, harassment, antitrust, and the like. Very little on the acquiring of genuine new skills. So that's, that's a tension. And we've got a tension. The spend is lower, the need is increasing. So the amount of training, but also the speed. I mean, the other thing that Joseph communicated was Technology is moving faster and faster. We've gotta get comfortable that the AI solution that we use this year might be out the window 2 years' time and be replaced by something else completely differently in 3 years' time. How are we going to manage this change in the workplace?

Not easily, I think is the short answer. So, you know, I start, when I think about this question, I start first with history. So if you look at any research about any kind of big disruption, like the one we're dealing with right now, The middle of it is very messy. There needs to be new technology, norms, rules, regulations to get through it. So that's first, recognize that. And second, you know, I think systems ultimately find their equilibrium. It's going to take time before we get there. I think workplaces right now are making some horrifically bad choices that feel good in the short term to hollow out apprenticeship training and development that they will regret in the medium and long term. So what then counterbalances that? I think you will see 2 things. I think one, the workplaces that take more of the long-term perspective and develop human capital that goes through apprenticeship and learns by experience, et cetera, will not only dramatically outperform those who don't, they will be a talent magnet and you will see massive separation. Number 2, I think new institutions will emerge to deal with this problem. So, you know, as by way of metaphor, 100 years ago, there's not really gyms. Why would you need a gym? It's an agrarian society. People are working 14 hours a day. They don't need to go and consciously work out because that's part of their day-to-day life. Today we have gyms where we can work on our muscles. I suspect somebody's gonna create a great business, which is a gym for the mind, which you're using less and less on the job, but you've got a new place where you can go and continue to learn and develop again in a fun way where you feel rewarded for it as opposed to being back in school. So that you can make sure that all the challenges that you and Joe talked about, which are very real, get addressed. I have not seen anyone with a business like this, but a gym for the mind, a gym for the soul, I think is a big opportunity waiting to be created. Last thing I'd say, you know, Singapore, as almost always, has gotten there faster than the rest of the world. I remember a few years ago, back when I lived there, before AI was even a big thing, Singapore was beginning to talk about completely changing the way that it viewed education. They said, historically it's been, we take an 18-year-old, we send them to school for 4 years or whatever, and they're done. Now we really have to view education like a subscription service where it is truly a lifelong thing where every once in a while you're gonna spike and you're gonna go somewhere for a year, 2 years, a week, 2 weeks, whatever it is. This idea that you're done at 21 or 22 is so antiquated, and this was true years ago. It's even more true today. So again, I think you're going to see new things emerging— education systems, third-party providers, organizations. But in the short term, I think it's going to be pretty ugly. It means all of us as professors, we got to up our game because our students are going to enter a world where there's going to be less support for them.

Some great ideas about the gym. And I totally get it that individuals will have to up their skill game, but we go to work in an organization called a business. Because we are interacting with other people in a team and because we've got skills to learn to create value in an organization. So even if we have individual training, we will still learn, need to learn skills to enable us to work together as a team. And I think what you're describing is, you know, you're not going to get this from a lecture. You're going to get this by some sort of experiential learning in a safe place to fail. And we call this play or games, right? But what would you say to someone says, well, play and games, that's not We don't need this in the business. That's for children. How would you reposition that thinking in their minds about this exploratory or experiential learning that's needed?

Well, what I would say is, do we accept the viewpoint that the world's changing faster than it ever has? Everyone will accept it. You can debate some of the nuances of that, but whatever, people agree. Would you accept the viewpoint that our people need to be able to adapt, need to change, need to do new things? Everyone would agree with that. Would you accept the viewpoint that change is easier when it's something that feels more fun and has less friction in it? Everyone would agree with that. So we have then a natural answer once we've done all of those things, which is, obviously, we're not talking about picking up crayons and coloring, although that has advantages too. We're saying, can we find something where there's an intersection? So can we bring work and fun together in a way where we are developing, we are stretching, we are learning, and We're also progressing against business objectives. You mentioned tensions before. We can frame this very easily as an either/or. We can say either we're working or we're having fun, or we can say there is a both/and possibility that when we bring these together, we can truly have our cake and eat it too, because we can take something that felt arduous and bring a lightness to it, which means not only are we doing it faster, we're doing it a lot better. So to me, it, it's not, let's make this a trade-off more. It's how do we find ways to bring these together? I agree. Most organizations are, are not ready for this. Most leaders would be like, what are you talking about? This sounds like kindergarten. So, you know, it'll take time. Darwinism, it works its way out. I, I really do believe organizations will embrace this and you'll see differences in performances, but we'll see. I could be wrong.

Are you seeing any examples of this in the business world?

Not in serious ways. You know, I mean, of course you've got, You've got isolated examples where people are creating, you know, the equivalent of labs teams, et cetera, where people have more freedom and more ability to go experiment, et cetera. But if there is an organization that has consciously and at scale brought the idea of integrating work, play, and fun together, I have not seen it. I'm sure there are organizations where, you know, they're in the gaming or whatever space where that's core, but a traditional organization that's done it, I think we're still waiting for the example. When you hear of one, please let me know because I would love to study it.

Well, I mean, I'm involved in some work where experiential games are in the workplace. For example, when you're trying to do cross-functional connecting supply and demand, I mean, you can describe this in a lecture, but the tension of knowing what it likes to feel like to be a demand planner is impossible to communicate unless you've experienced it yourself. And so if you're a supply planner interacting with that person, having some safe opportunity to experience the other people's roles makes it much better, I think, to understand each other and therefore be more productive as a result.

Hey, you know, Mark, you have actually that triggers a thought in my head. You know, you say, what are organizations that have done experiential development and have gamified it? The military, I think, would probably be the leading adopter of this, the US military in particular, where whether it's war games, simulations, models, et cetera, Always to say, I, I don't know if you describe any of those precisely as fun, but always to say we are going to try in very sophisticated ways to get people to experience something that is horrific so that they develop the ability to, in that actual circumstance, be able to respond appropriately. So I, I think that would be an organization that has really pushed the frontier for experiential learning. I would also say there are people who, you know, who provide solutions around this. There's a, a gentleman in Norway, Chris Rangan, who has a game called Transform, where it's a multiplayer board-based game that, like you described, someone experiencing what it's like to be in a function, it simulates what it's like to go through a transformation with activist investors coming at you and weird things happening with different companies and all that. And he has been very successful in bringing that to a range of organizations to help them again experience something that you only will episodically encounter in a business career. And that's one of the problems when something happens, you know, once in a career, It's hard to be prepared for it. Games and simulations are a great way to help with that.

Well, I think going back to Joseph Fuller, he gave another example, and I think the way we need to look at this is think about high-stakes situations. The military practices because the risk of getting it wrong is incredibly costly. Another example that Joseph Fuller gave was a CEO preparing for a pitch to investors or the quarterly investments report. He coached the CEO by playing a game. What they did is they trained AI agents in the roles of known investment bankers who regularly ask questions to that organization and its peer group competitors so that they could simulate how, in other words, play a game, how they might Uh, the, the analyst might ask questions to the CEO and prepare him in advance. If that's not gameplay, what is it?

No, absolutely. I love it. Great example. Joe, he, like me, had a first career in consulting. So I think maybe there's a little bit to those who have experienced it before they teach it that are willing to embrace and try different things. So though there's lots of academics that are trying different things, so I should not seem to be disparaging any of my more traditionally minded colleagues. They're doing awesome things as well.

So, that's in terms of wrapping up today's discussion. I'm sure we've got some listeners who have been inspired by this, but need to think, what am I going to do on Monday morning, either with my team or what are the conversations I'd like to start broaching with my boss about what we should do differently to get our teams upskilled and built the muscle to fight new challenges? Any thoughts?

Well, you know, I think the example that you gave from Joe is a great example, which is, What is a low-stakes way that we can go and engage with existing AI systems without requiring big new build or anything like that to go and simulate something? So what is a way we can take a problem that is persistent, that we face a bunch of times, and we can try a different way to develop and learn around it using existing AI tools? It's something, you know, I mentioned that the horizontal layer bringing AI into classes. One very simple thing I did the last time I taught my class, Leading Disruptive Change, is I designed a very rudimental simulation that was called the CEO Dilemma, where they have a choice between investing in today or investing in tomorrow. The obvious takeaway is if you do 100% in today, that's risky. You do 100% in tomorrow, that's risky. You've gotta balance it in the right way. But rather than having me lecture around it, students can go and get fired and they can go and try different things and see what works and what doesn't. So the thing that you could do Monday morning, I mean, I, I'm not a coder. It was 15 minutes of development to do it. I knew the question to ask. And I knew the way to frame things up, so it was very easy to develop it. What is the equivalent that you can do for your team? A, a problem that you want to make more visceral, an opportunity you wanna make more compelling, something where they can practice in a lower stakes environment. Those are the categories of things where experiential role-playing gamification can be really powerful. And one of the awesome things about AI is that which used to be complicated is now really simple. Classic disruptive innovation. Yeah. Which can completely change the way that we think about development if we do it in the right way.

Thank you. Thank you very much, Scott. That's been great and inspiring as ever. Well done. And I hope we can inspire others to make some change. Since I want to congratulate you since we last spoke, by the way, for all the great praise that Epic Disruptions has been getting. I saw it had been voted as one of the key books to read by Thinkers50, if I'm not wrong, and many other awards since. So, congratulations for that.

Thank you very much. Have you got

anything new coming up that people should look out for, or have we got to wait in anticipation for your next book?

Well, there's 2 answers to that question. First, I contributed a chapter to a compendium that's coming out in July. The book is called Human-Centered Leadership. It's done by the Silicon Valley Guild in conjunction with Thinkers50, 2 organizations that I have connections to. So my chapter is about the extraordinary possibilities of today's time. It's an optimistic optimistic chapter about what we can do in the midst of the great unfreezing. And I'm just about to start working on the next book, which I'm hoping comes out in the fall of 2027. The provisional title echoing off of Epic Disruptions is Epic Leadership: What to Do When Disruption Goes On and On and On and On. So that you have to wait a little longer for that one. But the compendium book is coming out this July.

Fabulous. Thank you. Well, if you can teach us how to keep an organization going from what you're saying to both explore and exploit, I'm going to be one of the first to buy the book. No doubt about that. Thank you very much, Scott.

Thank you, Mark. It's been fun.

Great. Bye-bye. This podcast is brought to you by Arcaro. To learn more about how Arcaro might help your organization adapt to a changing world, please visit our website, www.arcaro.com. More information in the show notes. Thank you.